I used to have a UMAX 600S SCSI scanner. I used it from 1997 thru to 2011. It was the best scanner I ever owned. The depth of field was perfect for scanning retro hardware PCB's. When it died, I got some HP ScanJet 7400C, which is USB + SCSI, however the scanning depth of field isn't that great, which shows on my scans. Look at that DTC VLB SCSI card I scanned. You can see which IC's are closer to the scanner glass and which are further based on how blurry they are. The 600S didn't have this issue.

I have a professional Linotype Hell scanner that I got from a media agency. Only had it tested on an old Mac so far, but it should work with Win9x, too.

PC hardware is odd sometimes... In an attempt to further optimize performance, I changed the L2 cache mode from WB to WT. WT should be slower, shouldn't it? Guess what..? Now Quake timedemo 3 gives me 14.2 fps instead of 13.8. 😁 Dr. Hardware shows better scores as well (>77000). Also funny, speedsys *always* shows 59.76, not matter at which settings. Even with turbo disabled, which halves the speed of the system, speedsys still shows a score of 59.76. I begin to think that it doesn't really measure that value but somehow guesses it based on the detected CPU, memory speed, etc.

Speedsys is a very poor indicator of CPU performance. I generally do not use it to compare small changes in system configuration.

For WT vs. WB L2 cache, I suggest using the Ziff-Davis CPUMark99 in Windows. It is their stand-alone product. I have found this benchmark to be the most sensitive to changes in L2 modes. I have attached this benchmark for your convenience.

Thanks, I'll try that when I'm back home later.

Thinking about it, the better performance with WT L2 cache makes sense. The AMD X5 already has WB L1 cache, so having L2 set to WB as well causes double-buffering of RAM writes, which isn't good. When a WB cache is flushed, all changes should be written to RAM. But if both caches are WB and you flush L1, changes are written to L2 instead, which must be flushed then as well, causing performance to suffer. So if L1 is WB, L2 should be WT. If L1 is WT, L2 should be WB. I'll double-check this later using a Cyrix 5x86, since the AMD's L1 mode can not be changed.

Spent some time trying to squeeze some more speed out of it by trying a FSB of 50 MHz and a multiplier of 3x. While video speed got faster by more than 30%, overall performance didn't get better. Even with 50ns RAM and 0 waitstates, I wasn't able to get the same memory throughput as with 4x 40 MHz. From all games, only DOOM got faster (752 realtics instead of 906). So I'm staying with 4x 40 MHz. At least I know now that my X5-133ADW is stable even with an FSB of 50 MHz. 😀

derSammler wrote:

So if L1 is WB, L2 should be WT. If L1 is WT, L2 should be WB. I'll double-check this later using a Cyrix 5x86

It would be interesting to quantify this with some benchmarks. I've always noticed that WB L2 cache improves speed very little. I've never seen it reduce performance much though.

Difference is very low. With the X5 at 160 MHz I got these results with Dr. Hardware yesterday:

* L1 WB, L2 WB: ~77100
* L1 WB, L2 WT: ~78300

And .4 fps improvement in Quake. So it's only about 2-3%. I'm pretty sure WB always give a speed boost when writing lots of data to RAM. But for reading from RAM, WT seems slightly faster, because of less overhead in the cache logic.

I'm going to benchmark all combinations with the Cyrix 5x86 later.

And the last free PCI slot goes to... this: 😁

derSammler wrote:

And the last free PCI slot goes to... this: 😁

I must have had the same card for years, and she worked for year after year until lightning killed three of my pc's at once 😠
The realmagic cards where plenty and the software worked well with them. Only issue I had was that the cable was a bit loose from the passthrough, and if you moved the tower a bit you had to reseat it to make it work again.

Tried today, but no dice. With the X5 running at 160 MHz, it is *just* fast enough to decode MPEG-1 using the card without stuttering (which is pointless, as it can do that in software only anyway). DVD is completely unusable. At least over IDE, there's no way for the CPU to push data from the DVD fast enough to the card. Too bad, really.

If you are interested in DVD playback on a socket 3, you might be interested in finding a Dxr2 decoder card. Re: DVD hardware decoder cards

...you can play DVD movies on a pci-based 486 using a Creative DXR2. My 486(s) have the UMC 8881/8886 chipset. I used a Cyrix 5x […]
Show full quote

...you can play DVD movies on a pci-based 486 using a Creative DXR2. My 486(s) have the UMC 8881/8886 chipset. I used a Cyrix 5x86-120 during this test.

I tested the card in Windows 98SE and Windows NT4.0 with a SCSI DVD-ROM and an IDE DVD-ROM. The SCSI DVD-ROM was SCSI2 and the IDE DVD-ROM was in PIO-4 mode using the onboard IDE connector (M919 motherboard).

In Windows 98SE, the resource meter showed about 44% cpu usage when playing a DVD with the IDE drive, whereas 20% with the SCSI drive. In Windows NT4.0, I only tested the SCSI drive and task manager did not show any cpu usage above that of an idle task manager. I had no issues with sound. My default resolution is at 1024x768, however I also tested DVD playback at 1280x1024 which also worked, but was not as clear. At 1280x1024, the VGA pass-thru isn't as crisp even when browsing around Windows, but not so bad that it would be noticed if using a CRT monitor.

From the incredibly low CPU usage, I wonder how slow a 486 CPU DVD playback will work with? I wonder why Creative set the system requirements to a Pentium 100; the pentium instruction set doesn't appear to be used in the Creative DVD Player's software.

At 1024x768, the DVD playback quality is quite good. If I playback at full screen, and sit back about a meter, I don't think I could tell the difference between that and software based playback on a P4+.

The decoder card is not the problem here, but bus speed. The Creative card uses the same EM8300 decoder chip anyway. Maybe Windows 95 can't handle all that as good as 98 or NT.

Also, a Cyrix 5x86 is not a 486.

feipoa wrote:

If you are interested in DVD playback on a socket 3, you might be interested in finding a Dxr2 decoder card. Re: DVD hardware decoder cards

What do you think of using an ATI 3D RAGE II+DVD in a socket 3 board?
What drivers do you recommend, what DVD playback software supports the hardware acceleration?

I have this card, but when I put it in the computer, blank screen. I'm wondering if it has an Apple Macintosh firmware. Anyone have the PC firmware for this card? My card could also be defective.

I will try to get the BIOS for you.
But it may take a few days...

Great, thanks. Few days or a few months, no rush.

Yeah, zergling rush. On a 486. 😀

Disruptor wrote:

What do you think of using an ATI 3D RAGE II+DVD in a socket 3 board?

The Rage II+DVD has DVD acceleration only, that is color-space conversion and motion compensation. It does not have a full-featured MPEG-2 decoder. There's no way to play DVDs using that card on a socket 3 system.

isnt motion comp the hardest part about that? but then again i suppose the compressed data stream alone can be too much for a 486 (or so i imagine)
what are the other steps that are missing here?